Today, popular Carrier Ethernet services being defined by the Metro Ethernet Forum (MEF, online at metroethernetforum.org) include E-Line (for Point-to-point), E-Tree (for point-to-multi-point) and E-LAN (for multi-point) configurations. Depending on how bandwidth is allocated (i.e. dedicated or shared), these services may be defined as Ethernet Private Line/LAN (dedicated bandwidth) or Ethernet Virtual Private Line/LAN (shared bandwidth). These services are growing in popularity and will form the basis of future private and public network connectivity. For an Ethernet Virtual Private Line (EVPL) service, point-to-point bandwidth is assigned at Layer 2 through the use of packet tagging with oversubscription allowed. EVPL services are offered at a range of data rates from a few Mbps to Gbps and are typically implemented over native Ethernet or Multiprotocol Label Switching (MPLS)/Virtual Private Wire Services (VPWS) technologies. Layer 2 switching and transmission resources are shared with other services on the network. In the case of an Ethernet Private Line (EPL) service, bandwidth is dedicated at Layer 1 or 0 using Time Division Multiplexing (TDM), Wavelength Division Multiplexing (WDM), or fiber to partition the service from other services. By dedicating bandwidth in this way, oversubscription is not possible. Instead, the full rate of the connection is allocated to the customer, whether used or not. EPL services are typically defined for GbE or 10 GbE point-to-point connections. They are implemented over Wavelength Division Multiplexed (WDM), Synchronous Optical Network (SONET), Synchronous Digital Hierarchy (SDH), and increasingly over Optical Transport Network (OTN) technologies. Layer 2 bandwidth is not shared, but Layer 1 switching and transmission resources may be shared with other services on the network.
An Ethernet Virtual Private LAN (EVPLAN) service is similar to the EVPL except that it supports more than two user endpoints in a LAN configuration. Again, oversubscription is allowed. EVPLAN services may be supported over native Ethernet or MPLS/Virtual Private LAN Service (VPLS) technologies. Layer 2 switching and transmission resources are shared with other services on the network. Of the aforementioned service types, the virtual EVPL and EVPLAN services are popular because they offer the network operator the opportunity to oversubscribe bandwidth providing efficient use of network resources. While in many respects it is advantageous to multiplex many packet services across a single packet infrastructure (e.g. using IP, MPLS or native Ethernet technologies), many customers require dedicated and private connectivity services. Consequently, dedicated EPL services are very popular with large enterprise and wholesale carrier market segments that require dedicated bandwidth to build or supplement their own networks. This market segment has a need for Ethernet private LAN (EPLAN) connectivity in addition to EPL. Today a number of approaches exist for Ethernet private LANs, such as, for example, operating separate physical Ethernet networks over different physical network topologies. This requires that dedicated, separate Ethernet switches are used for each Ethernet private LAN service and connectivity to those switches is provided over EPL links. Unfortunately, this implementation is counter to the ongoing desire for convergence and consequently can be operationally challenging and expensive to deploy.
Alternatively, an approach may include operating separate Ethernet network instances using Virtual LAN (VLAN) or Service Instance Identifier (I-SID) differentiation on a common Ethernet infrastructure. This approach does not provide the full degree of partitioning provided in the previous example, but resources can be reserved in the Layer 2 network and dedicated to the Ethernet private line service. As an Ethernet bridged network, this approach is advantageous in that the service bandwidth demands scale linearly with the number of user endpoints (N). However, it is fundamentally a shared Layer 2 implementation. Therefore, to make sure that all sites offer the potential to act as an add/drop location (or a User-Network Interface (UNI)), all Ethernet bridges must participate in a single network topology (within which specific service instances are defined). The topology is organized using a spanning tree protocol (or, in the case of Shortest Path Bridging (SPB), a routing protocol) to define a loop-free forwarding topology. Then, for any given single service instance only a subset of the Ethernet bridges are actually used as UNIs, with the remainder acting as tandem forwarding devices. In many network locations (especially for large networks), the tandem traffic through a bridge can be large and can result in inefficient use of the packet fabric. In such situations, where Layer 2 forwarding decisions are not really required (e.g. degree-2 sites), it would be beneficial to bypass the packet fabric completely and so free up its switch capacity for additional new service instances (this is a similar problem to the much publicized ‘IP router bypass’ challenge). This situation becomes particularly evident when a large bandwidth user's VPN shares the same network as multiple small bandwidth VPNs. Unfortunately, the creation of a bypass link in an Ethernet network is not practical as it creates a new Layer 2 topology resulting in potential loops, thus requiring the re-definition of a new loop-free tree.
Yet further, an approach may include operating separate Ethernet network instances across separate MPLS or VPLS connections. This can be costly due to the higher cost per bit of IP/MPLS devices (relative to Ethernet switches). In addition to the transit issue described previously, MPLS/VPLS suffers from an N2 bandwidth scaling inefficiency. Each of the above is not ideal for the private bandwidth customer either due to cost or lack of trust in the shared approaches. Instead of using the above methods, many customers will choose to build their own private networks using multiple EPLs connecting their own switches together in a mesh configuration. This results in an N2 connectivity inefficiency and the added operations complexity of operating their own WAN switches.